Hutchins’ anthropological study of a team involved in the navigation of a Navy ship reveals the manner in which complex computational tasks are executed via a social organization. His observations of the way cognitive computation differs at the agent-level vs. team-level relates closely to the way groupwork in SharedPlans is designed to preserve collective intentionality while also assuring efficiency in task allocation and completion.
The hierarchical structure in both SharedPlans and on Hutchins’s ship seems to still leave unanswered the question: “What entails essential knowledge?” Hutchins emphasizes the importance of proper division of labor, yet also shows repeatedly the ways in which redundant knowledge, collective memory, and modularity are crucial to the emergent properties of group cognition.
How can an AI framework such as SharedPlans both install efficiency in sub-task allocation (allowing for parallelism in execution as well as limiting the amount of knowledge that must be stored by individuals/sub-groups) while incorporating Hutchins’s insights in the possibly "inefficient" properties that make group cognition qualitatively different from the sum of individual capabilities?